"Daily Digest": An Expanded Vision
Concept Philosophy: Turning Your Digital Footprint into a Story
We spend countless hours on our computers, but our digital memories are fragmented across browser histories, file modification dates, and application logs. This data is cold, raw, and meaningless. "Daily Digest" is built on a simple, powerful idea: What if an AI could act as your personal, private chronicler, transforming this chaotic data into a meaningful narrative of your day?
It’s not a surveillance tool or a productivity tracker in the traditional sense. It's a mindfulness application designed for zero-effort self-reflection. It helps you notice patterns in your focus, creativity, and relaxation, providing a story of your digital life that would otherwise be lost.
Detailed Feature Breakdown
The Core Narrative Summary:
At the end of each day (or a user-defined time), the app synthesizes your activity into a 200-300 word summary.
Intelligent Highlighting: It doesn't just list events. The LLM is prompted to identify the day's primary themes, major tasks, and notable shifts in activity.
Observant, Non-Judgmental Tone: The language is crafted to be reflective, not critical. It says "You spent the afternoon exploring articles on ancient history" not "You were distracted for 4 hours."
Topic & Project Clustering:
The AI automatically identifies and groups related activities. Instead of listing 20 separate file edits and website visits, it summarizes: "A significant part of your morning was dedicated to the 'Q4 Marketing Campaign,' involving work on the main presentation, the budget spreadsheet, and several emails to the design team."
Inferred Mood & Focus Analysis:
By analyzing the speed of switching between tasks, the types of applications used, and the content browsed, the LLM can add a layer of insightful analysis.
Example Output: "The day began with a period of intense focus, with most of your time spent in a single coding environment. The afternoon saw a shift towards broader research and communication across multiple applications."
Searchable, Encrypted History:
All past digests are saved in a local, encrypted database.
The user can perform natural language searches like: "When did I work on the 'Andromeda' project?" or "Find the day I was researching flights to Japan."
Customizable Privacy Controls:
During setup, the user can explicitly choose which applications and websites the agent is allowed to observe. They can blacklist sensitive apps (like password managers) or websites to ensure total privacy.
The User Experience (UX/UI)
The Agent: A tiny, unobtrusive icon in the system's menu bar or tray. It provides zero distractions during the day. Its only job is to collect data silently and privately.
The Onboarding: The first launch is critical. It will clearly explain what data is collected, how it's used, and that it never leaves the computer. It will then guide the user through setting permissions and privacy preferences.
The Digest View: The main interface is a clean, minimalist reader resembling a high-end journaling app. You can view today's digest or browse past entries through a calendar view. The focus is on beautiful typography and a calm, reflective experience.
Deep Dive: Technical Architecture
The Data Collector (The Agent):
This is a lightweight background process built using platform-specific tools (e.g., Swift/Objective-C on macOS, C# on Windows) to access Accessibility and OS APIs.
It logs high-level events (app focus changes, window titles, website URLs/titles) into a simple, structured JSON log file. Example: {"timestamp": "...", "app": "Chrome", "title": "Ollama - Run LLMs Locally"}.
The Brain (Local LLM Engine):
Ollama is used to manage and run a local gpt-oss model (e.g., a quantized version of Llama 3 Instruct or Mistral 7B Instruct).
The "Secret Sauce" Prompt: The daily JSON log is passed to the LLM with a highly engineered prompt.
Example Snippet: "You are 'Digest,' a personal chronicler AI. Your task is to transform this raw log of computer activity into a concise, insightful, and well-written narrative summary of the user's day. Identify 2-3 main themes or projects. Group related activities together. Adopt a warm, observant, and non-judgmental tone. Do not simply list the events; synthesize them into a story. Here is the log: [...]"
The Interface (Frontend):
Built with a cross-platform framework like Tauri or Electron for rapid development. It reads the generated digests from a local database and presents them in the clean UI.
Data Storage:
A single, encrypted SQLite database file stores everything: the raw logs, the generated digests, and user preferences. This ensures data is portable and secure.
Log in or sign up for Devpost to join the conversation.